Elon Musk’s AI company, xAI, late on Wednesday released its latest flagship AI model, Grok 4, and unveiled a new $300-per-month AI subscription plan, SuperGrok Heavy.
Grok is xAI’s answer to models like OpenAI’s ChatGPT and Google’s Gemini, and can analyze images and respond to questions. In recent months, Grok has become more deeply integrated into Musk’s social network, X, which was recently acquired by xAI. However, that has also put Grok’s misbehavior front and center for millions of users.
The expectations are high for Grok 4. The latest AI model from xAI will be stacked up against OpenAI’s forthcoming AI model, GPT-5, which is expected to launch later this summer.
“With respect to academic questions, Grok 4 is better than PhD level in every subject, no exceptions,” said Elon Musk during a livestream Wednesday night. “At times, it may lack common sense, and it has not yet invented new technologies or discovered new physics, but that is just a matter of time.”
The launch of Grok 4 comes amid a tumultuous week for Elon Musk’s companies. Earlier on Wednesday, Linda Yaccarino stepped down from her role as the CEO of X after roughly two years with the company. X has yet to announce her successor.
Yaccarino’s departure comes just days after Grok’s official, automated X account responded to users with antisemitic comments criticizing Hollywood’s “Jewish executives” and praising Hitler. xAI had to briefly limit Grok’s account and delete the offensive posts. In response to the incident, xAI appeared to have removed a recently added section from Grok’s public system prompt, a list of instructions for the AI chatbot to follow, that told it not to shy away from making “politically incorrect” claims.
Musk and xAI’s leaders largely avoided discussing the incident, instead focusing on Grok 4’s performance and capabilities.
xAI launched two models on Wednesday: Grok 4 and Grok 4 Heavy — the latter being the company’s “multi-agent version” that offers increased performance.
xAI claims that Grok 4 shows frontier level performance on several benchmarks, including Humanity’s Last Exam— a challenging test measuring AI’s ability to answer thousands of crowdsourced questions on subjects like math, humanities, and natural science. According to xAI, Grok 4 scored 25.4% on Humanity’s Last Exam without “tools,” outperforming Google’s Gemini 2.5 Pro, which scored 21.6%, and OpenAI’s o3 (high), which scored 21%.
xAI claims that Grok 4 Heavy, with “tools,” was able to achieve a score of 44.4%, outperforming Gemini 2.5 Pro with tools, which scored 26.9%.
The nonprofit Arc Prize says that Grok achieves a new state-of-the-art score on its ARC-AGI-2 test — another difficult benchmark that consists of puzzle-like problems where an AI has to identify visual patterns — scoring 16.2%. That’s nearly twice the score of the next best commercial AI model, Claude Opus 4.
Alongside Grok 4 and Grok 4 Heavy, xAI launched its most expensive AI subscription plan yet, a $300-per-month subscription called SuperGrok Heavy. Subscribers to the plan will get an early preview to Grok 4 Heavy, as well as early access to new features. The plan is similar to ultra-premium tiers offered by OpenAI, Google, and Anthropic, but xAI now offers the most expensive subscription among major AI providers.
SuperGrok Heavy subscribers may get early access to some new products xAI plans to launch in the coming months. The company said Wednesday that an AI coding model is coming in August, a multi-modal agent in September, and a video generation model in October.
xAI is releasing Grok 4 through its API in an effort to get developers to build applications with the model. The company notes that xAI’s enterprise sector is only two months old, however, it plans to work with hyperscalers to make Grok available through their cloud platforms.
Despite Grok’s frontier-level performance on benchmarks, it may prove difficult for xAI to move past its recent mishaps as it tries to pitch Grok to businesses as a real contender to ChatGPT, Claude, and Gemini. Whether businesses are ready to adopt Grok, flaws and all, remains to be seen.
